Cached input
Repeated prompt prefixes are billed at a reduced input rate. RunAPI passes the discount through at 50% of OpenAI's cached rate.
LLM API Pricing
OpenAI API Pricing in 2026
GPT-5.4 costs $2.50 input and $15 output per million tokens; GPT-5.5 costs $5 and $30. RunAPI gives you the same models at half those rates — same API, same output, same behavior, lower number on your invoice.
Updated June 18, 2026
RunAPI Editorial
At a glance
What does the OpenAI API cost right now?
OpenAI bills per million tokens, with separate rates for input (your prompt) and output (the model's response). Output tokens cost more — the model has to generate each one. There is also a cheaper cached-input rate when you resend the same prompt prefix. All figures below are per million tokens.
Most popular
GPT-5.4 at $1.25/M input and $7.50/M output through RunAPI. Official rate is $2.50/$15.
Cheapest option
GPT-5.4-mini — the cheaper option for high-volume work where you don't need the full flagship. Billed at 50% off through RunAPI.
Maximum capability
GPT-5.5 at $2.50/M input and $15/M output through RunAPI. Official rate is $5/$30.
Cache discount
Cached input tokens cost a fraction of standard input — passed through at 50% on RunAPI.
Model-by-model breakdown
How much does each GPT model cost per million tokens?
The table shows official OpenAI pricing alongside RunAPI pricing. Output tokens cost significantly more than input tokens — that is by design, since generating each token requires more compute than reading it. RunAPI applies a flat 50% discount across all GPT models. No volume commits, no subscriptions.
| Model | Official input /M | Official output /M | RunAPI input /M | RunAPI output /M | Context window |
|---|---|---|---|---|---|
| GPT-5.5 | $5.00 | $30.00 | $2.50 | $15.00 | 400K |
| GPT-5.4 | $2.50 | $15.00 | $1.25 | $7.50 | 400K |
| GPT-5.4-mini | $0.25 | $2.00 | $0.13 | $1.00 | 400K |
| GPT-5.3-codex | $2.50 | $15.00 | $1.25 | $7.50 | 400K |
Cache and batch
How do cache and batch discounts cut your GPT bill?
OpenAI bills less for cached input tokens and cuts batch request costs in half for jobs that can wait up to 24 hours. Both matter if you're running a coding agent or processing large batches — repeated context and bulk work add up fast.
Batch API (50% off)
Requests submitted to the Batch API run at half the standard rate with up to 24-hour turnaround. RunAPI passes this through on top of its own discount.
Reasoning effort
GPT-5 models let you set reasoning effort. Lower effort emits fewer reasoning tokens, directly reducing output cost on metered billing.
Output token control
Cap max output tokens per request to bound cost and avoid runaway generations on long agentic tasks.
Real-world costs
What does the GPT API cost for real workloads?
Token rates look abstract until you map them to real work. A customer-support chatbot handling 50 conversations a day is a different cost profile than a coding agent running all day or a content pipeline processing hundreds of documents. Below are common developer workloads — including automation workflows — with estimated monthly costs at two usage levels, billed at RunAPI rates.
| Workload | Model | Light use (~50 tasks/day) | Heavy use (~200 tasks/day) | Monthly saving vs official |
|---|---|---|---|---|
| Coding agent (Codex) | GPT-5.3-codex | $20/mo | $80/mo | $20–$80 |
| Customer-support chatbot | GPT-5.4-mini | $6/mo | $24/mo | $6–$24 |
| RAG knowledge assistant | GPT-5.4 | $18/mo | $72/mo | $18–$72 |
| Content generation pipeline | GPT-5.4 | $25/mo | $100/mo | $25–$100 |
| Multi-agent orchestrator | GPT-5.5 | $90/mo | $360/mo | $90–$360 |
Provider comparison
Is the OpenAI API cheaper than Claude and Gemini?
Locking into a single provider is a real risk when one price cut overnight can make a competitor 30% cheaper. Here is how the flagship models compare per million tokens — OpenAI, Anthropic, and Google side by side — so you can make the call without digging through three billing pages. RunAPI gives you one API key that routes to all three.
| Provider | Flagship model | Input /M | Output /M | RunAPI rate |
|---|---|---|---|---|
| OpenAI | GPT-5.4 | $2.50 | $15.00 | $1.25 / $7.50 |
| Anthropic | Claude Opus 4.7 | $10.00 | $50.00 | $5.00 / $25.00 |
| Gemini 2.5 Pro | $1.25 | $10.00 | $0.63 / $5.00 |
RunAPI applies a 50% discount on all providers listed above. Prices verified June 2026.
Getting started
How to access the GPT API through RunAPI
1
Create a RunAPI account
Sign up at runapi.ai. No credit card required for the free tier.
2
Copy your API key
Go to Dashboard → API Keys. Create a key and save it — you will use this as your OpenAI API key.
3
Point your SDK to RunAPI
Set the base URL to https://api.runapi.ai/v1 and use your RunAPI key. Any OpenAI-compatible client works.
4
Start making requests
Use gpt-5.4, gpt-5.5, or any GPT model ID in the model parameter. RunAPI handles routing and billing at 50% of the official rate.
Frequently asked questions
OpenAI API Pricing FAQ
How much does the OpenAI GPT-5 API cost?
GPT-5.4 costs $2.50 per million input tokens and $15 per million output tokens officially. GPT-5.5 costs $5 and $30. Through RunAPI, every GPT model is billed at half those rates — GPT-5.4 runs $1.25 input and $7.50 output per million tokens.
Why is RunAPI 50% cheaper than OpenAI?
RunAPI negotiates volume pricing with model providers and passes the savings on to developers. Requests reach the same OpenAI models with identical output, safety filters, and behavior, so the only difference is the lower rate on your invoice. There is no quality trade-off and no separate billing tier — the discount applies automatically to every GPT model.
Is there a free tier for the OpenAI API?
Yes. New RunAPI accounts get free credits to test any GPT model — no credit card needed. After that, billing is pay-as-you-go with no monthly minimum and no subscription. Fund a balance and each API call deducts its token cost. You can top up any amount and track usage per model in the dashboard.
I'm building an n8n bot — how does API billing work for automation workflows?
Yes, and it works well for automation tools like n8n, Make, or Zapier. Each workflow step that calls a GPT model is billed by tokens used — there is no separate per-call fee or rate limit tier. For workflows that run on a schedule or process batches, the Batch API discount (50% off) applies and stacks with RunAPI's base discount, so recurring automation jobs cost significantly less.
How are image and audio inputs priced — it's not just tokens?
Image inputs are billed by tile — OpenAI divides the image into 512×512 tiles and charges per tile. Audio is billed per second of input. Neither uses the per-million-token unit. RunAPI passes through image and audio pricing at the same 50% discount, so a GPT-4o vision request or Whisper transcription costs half the official rate.
Can I use the OpenAI SDK with RunAPI?
Yes. RunAPI is OpenAI-compatible. Point any OpenAI client at https://api.runapi.ai/v1, use your RunAPI key, and pass a GPT model ID. Existing code that already uses the OpenAI SDK works without any changes beyond the base URL and key, so migrating an established project takes about a minute.
Does RunAPI support GPT-5.3-codex for coding?
Yes. GPT-5.3-codex is available through RunAPI at 50% of the official rate, which is $1.25 input and $7.50 output per million tokens. It works with Codex and other OpenAI-compatible coding tools by overriding the base URL and key in their settings. Cached input and batch discounts also pass through, lowering the effective cost of repetitive coding sessions further.
RELATED
Explore More
Claude API Pricing
Claude Opus, Sonnet, and Haiku pricing at 50% off Anthropic rates.
ChatGPT API Pricing
ChatGPT and GPT-5 pricing comparison for developers.
Codex Pricing
What Codex CLI costs per session with RunAPI billing.
All Model Pricing
Compare 130+ models across text, image, video, and audio.
Run GPT-5 at half price.
Create a free RunAPI account, get your API key, and call any OpenAI GPT model at 50% off official pricing.